Use your own AI model
Add your own Claude, ChatGPT or OpenAI-compatible account under Settings so answers use it first, and see exactly what still uses Breno's own models.
By default Breno answers with its own models and spends your workspace's coins. If you would rather use your own account, add it: usage then bills to your provider, and Breno tries your account before workspace credits.
Add an account#
Open Settings, then Your AI accounts and add an account. Choose a provider:
| Provider | What it is |
|---|---|
| Claude | Your own Claude account, with a Claude API key. Usage bills to your Anthropic account. |
| ChatGPT | Your own OpenAI account, with an API key. Usage bills to your OpenAI account. |
| Other API key | Any OpenAI-compatible endpoint: Gemini, Mistral, Groq, Together, Azure OpenAI or your own gateway. You give a Base URL and a Model ID. |
Paste the key. You can give the account a Label and, for Claude and ChatGPT, choose a Model; left empty, the provider's default is used.
How accounts are used#
Breno starts with the first enabled account. When an account fails or hits its limit, it moves to the next, and then to workspace credits. Use the move up and move down controls to set the order, and the switch on each account to turn it on or off. Remove stops Breno using a key immediately.
What it changes, and what it does not#
Workspace default#
Separately, an administrator can choose the workspace's default model provider (for example Qwen, GLM, ChatGPT or a Gemini-compatible endpoint) in the workspace's AI settings. That default is what is used when no personal account is available.
Keys are stored encrypted#
Keys you add are encrypted at rest and are never returned to the browser; only a short hint is shown. Production deployments refuse to start without the encryption key configured. See Security overview.